Research On Surface Defect Detection Of Aluminum Tube Based On Machine Vision

Aluminum tube is a commonly used tube,which is widely used in construction industry and automotive industry etc.In automotive tubular radiator aluminum tube is the cooling tube,which is an important part of automotive radiator.Aluminum tubes may be damaged during production and transportation,so they need to be inspected for defects and sorted before they are assembled with the heat sink into a radiator core.Currently,the task of detecting defects in aluminum tubes is mainly achieved by manual inspection,but the efficiency of human eye in ranking the quality of aluminum tubes is low and difficult to meet the needs of industrial production.In this paper,machine vision technology is used to inspect the surface defects of aluminum tubes to make up for the shortcomings of manual inspection by taking advantage of the efficiency and repeatability of machine vision.The paper analyzes the types of defects on the surface of aluminum tubes and their causes,designs the image acquisition device for the detection of defects on the surface of aluminum tubes according to the industrial process requirements,and determines the workflow of the defect detection unit and the hardware selection.Because the surface of aluminum tubes has reflective properties to light,the dark field of view illumination is used.First,the acquired image is pre-processed,and for the phenomenon of uneven light distribution and high light saturation in the aluminum tube image acquisition,this paper proposes an image enhancement algorithm based on double Gaussian filtering,which uses a combination of numerical similarity Gaussian function and multi-scale Gaussian function to extract the light component of the image,and recover the correction of the light component to effectively improve the area of uneven light in the aluminum tube image.The edge detection algorithm locates the aluminum tube region and removes the background image to prevent the background region in the image from interfering with the information processing of the aluminum tube part.Secondly,in order to achieve higher accuracy detection of good and defective aluminum tubes,this paper proposes a robust principal component analysis(RPCA)-based algorithm for detecting surface defects of aluminum tubes,which obtains sparse images by solving the RPCA model of the original data matrix of aluminum tube images,threshold segmentation of sparse images and morphological processing,and determine whether there is a defect in the aluminum tube based on the comparison results of the defect pixel area of the binary image and the preset threshold.Finally,this paper adopts the deep learning method to classify and recognize the aluminum tube images with defects,and first makes the aluminum tube defective image dataset,then trains and tests the model by adjusting the Alexnet network parameters,and verifies the performance of the network model to classify the aluminum tube defective images according to the accuracy of the test set.The detection accuracy of RPCA-based aluminum tube surface defect detection algorithm for aluminum tubes can reach 97.3%,which proves the effectiveness of the algorithm,but the algorithm does not identify the defect category of aluminum tubes,so the Alexnet network-based method is used to classify the aluminum tube surface defect images,and the experimental results show that the accuracy of the defect image classification of the method can reach 92.91%,which meets the industrial requirements for classification accuracy.